[ros-users] slam_gmapping: extracting pose history

[Brian Gerkey]

On Mon, Apr 26, 2010 at 10:26 AM, Ivan Dryanovski
<ivan.dryanovski at gmail.com> wrote:
> I have the slam_gmapping package up an running. I am trying to extract
> the full, slam-corrected pose history of the robot at any given point
> during the mapping process. I know that gmapping node publishes a map
> → odom transform. My initial guess was to just take the transorms in
> chronological order to reconstruct the pose history. However, to my
> understanding, the published transform is for the current best
> particle. When a resampling occurs and the current best particle
> changes, the transforms published so far will become irrelevant. Is
> there any way to get around that, or am I misinterpreting the way that
> the gmapping package works?

hi Ivan,

You're correct that, at each iteration, the slam_gmapping node selects
the highest-weight particle from the filter and converts its current
pose to a map->odom transform.

It sounds like you're looking for the full pose history of the current
best particle.  I believe that the underlying GMapping library
maintains this information, and it should be pretty easy to modify
slam_gmapping to expose the information to ROS.   I'd happily accept a
patch adding this feature; if it were to arrive very soon (next couple
of days), it could be included in slam_gmapping 1.2.

	brian.